Since, the results of Relax refinement doesn't make much differences.
Thank you very much!
If I understand your question correctly, you're saying that you've tried using the relax application, are unsatisfied with the diversity of structures that it produces, and are wondering if there are protocols you can use which will sample conformations further from the starting structure. (If that's an inaccurate summary, you might want to elaborate on what you mean by "relax ... doesn't make much difference".)
There's several options. If you have a particular region you want to do more extensive sampling in, there are a number of remodeling protocols, most of them oriented around loops (although many can handle larger segments with a variety of secondary structure). There's the regular loop modeling program (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/d1/d...), a fragment based loop modeling program (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/app_...), and one that's based on kinematic loop closure (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/df/d...). There's also a localized backbone movement based on the "backrub" move (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/d1/d...).
If you want a more global perturbation, and have structures of related proteins, you can try comparative modeling (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/d5/d...) to generate structures which are more-or-less structurally similar to your desired protein, but might incorporate features from related proteins.
If you have a bunch of computing power at your disposal, and the previous protocols aren't extensive enough sampling, you can try the ab initio folding application (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/app_...).
Depending on you structure, there may be more specialized applications, like fold and dock (http://www.rosettacommons.org/manuals/archive/rosetta3.3_user_guide/d0/d...) for ab initio folding of symmetric protiens.